- Note the basic concepts in data classification.
- Discuss the general framework for classification.
- What is a decision tree and decision tree modifier? Note the importance.
- What is a hyper-parameter?
- Note the pitfalls of model selection and evaluation
1. Note the basic concepts in data classification
The classification of data include:
Data mining tasks are broadly classified into two categories:
i. Tasks that need prediction.
The goal of these tasks is to forecast the value of one characteristic depends on the
principles of other characteristics. The feature to be predicted is often referred to as the targeted
or predictor variables. The qualities utilized to make the prediction are referred to as the
analytical or independent variables.
ii. Tasks requiring description.
In this regard, the aim is to derive patterns, trends, groupings, trajectory, and
anomalies—which characterize the correlations that underlie data. Descriptive data mining jobs
are usually exploratory, necessitating post processing methods to validate and explain the
2. Discuss the general framework for classification.
Predictive modeling refers to developing a model for the target variable following the
explicative factors. Predictive modeling tasks have two types: categorization utilized for
continuous dependent variable and extrapolation for the discrete dependent variable. For
instance, it is a categorization job to forecast if a Website owner buys from an online library, as
the dependent variable is binary valued. In contrast, predicting a stock’s future price is a regression work since the price is a property that is constantly appraised. The objective of both
challenges will be to train a model that closes the gap between the target variable’s predicted and actual values. Clients who will react to a marketing strategy, forecast changes in the Earth’s
environment, or assess if a patient has a particular disorder based on medical testing data can be
used to discover predictive modeling.
3. What is a decision tree and decision tree modifier? Note the importance.
A decision tree is a configuration composed of a root node, branching, and leaf nodes.
Each internal node represents a test on an attribute, each branch represents the result of a test,
and each leaf node represents a class label. The root node is the trees highest node. Decision
Tree Categorization produces information in the form of a binary tree-like framework, allowing
marketing personnel to easily comprehend the results and identify relevant variables for churn
control. The rules of a Decision Tree classifier are used to forecast the dependent variable. The
Tree Classification technique gives a simple explanation of the data’s covariance matrix.
In comparison with other approaches, building decision trees may be very rapid. Another
benefit is the simple and easy to comprehend model decision, tree models. A decision tree is a
class parameter that split training reciprocally until each score contains instances from one class
in their whole or predominance. Therefore, each node of the tree containing the non-leaf split
point is a test of one or more properties.
4. What is a hyper-parameter?
Hyper-parameters are simulation variables that, without actual observable data, are
calculated. It is fundamentally a reasonable theory, without the real data, on what parameters a model may have. A hyper-parameter in data mining refers to a primary variable to be modified
for optimization. The k in the k-nearest neighboring algorithms is an illustration of such a
variable. These settings should only be adjusted in the data images without examining the real
data, resulting in bias.
5. Note the pitfalls of model selection and evaluation
In distributed computer settings, real-world data is generally saved on separate platforms.
It might be in databases, systems, or even on the Internet. Due to organizing and technological
considerations, it is almost impossible to bring all data into a single data repository. For instance,
several regional offices might have own servers to store their data, but all data (hundreds of
thousands of gigabyte) in all offices will not be stored on a single server. Data mining, therefore,
requires the development of models and algorithms that allow dispersed data to be mined.